Guides · Manage

IT spend benchmark: how to compare honestly

Published IT spend ratios disagree, and a single percentage can justify any budget. This guide shows how to build a benchmark that survives the CFO: a fixed scope, nine questions for every published figure, a confidence level for every comparison, and the views that say more than the ratio.

Why one number misleads

Sooner or later a board member asks how much the company spends on IT compared with others, and someone quotes a percentage of revenue from a survey. Taken alone, that number can justify a cut or an increase with equal confidence, and both are often wrong. Published figures for IT spend as a share of revenue can differ widely between sources, even for the same industry, because they measure different things:

  • Different scopes. Some count only the IT department's budget, others all technology spend, including what business units buy themselves.
  • Capital or depreciation. Surveys of IT budgets usually count capital spend in the year it is spent; the accounts show depreciation instead.
  • Average or median. An average is pulled up by a few heavy spenders. A median or a range says more.
  • Sample and date. Size, region and the job of the respondent all shift the answer, and ratios move over time.
  • Industry. Industry changes the answer more than anything else. A cross-industry figure says little about a manufacturer or a bank.

A benchmark is useful when the scope is clear, the sources are comparable and every figure can be traced. Then it tells you where to ask questions. It never sets a budget on its own. That is why this guide contains no league table of percentages, by design: it shows how to build a comparison you can defend.

Fix the scope with finance first

Before calculating any ratio, agree with finance what counts as IT spend. Three scopes cover most needs:

ScopeWhat it includesUse it to
NarrowThe IT budget: operating and capital spend in the year, including IT staff, contractors, hardware, software, cloud, outsourcing and telecomsCompare with most published surveys of IT spend
BroadNarrow, plus technology bought and run by business units: tools, local support, analytics contractorsShow the real total, and how much sits outside IT
Profit and lossTechnology cost as it appears in the accounts, with depreciation instead of capital spendReconcile with the accounts and the CFO's view

Write down the rules for the difficult items once: capital spend, depreciation, telecoms, IT staff in business units, shared services, and technology embedded in products. The same rules then apply every year.

Normalise before you compare

Even on the same scope, a few items distort any comparison. Adjust for them, or at least name them next to the result:

  • Outsourcing moves staff cost into services: compare total spend, not staff and external spend separately.
  • Capitalised internal labour moves cost out of operating expense: compare on the spend view, before capitalisation.
  • Acquisitions and disposals change both IT spend and revenue mid-year: use a pro forma year.
  • Currency: convert at one average rate for the year, and keep it for the trend.
  • The denominator: banks and insurers usually compare on operating income, not gross revenue; distributors with high pass-through costs look cheap on revenue whatever they spend.

Nine questions to ask of any published figure

Before quoting a figure to the CFO or the board, ask these. If you cannot answer most of them, leave the figure out.

QuestionWhy it matters
Who published it, and why?Vendors and advisers publish surveys to start conversations. Read their samples closely.
When was the data collected?Spend ratios move. A five-year-old figure describes another market.
How many respondents, and who?Size, region and the job of the respondent all bias the answer.
Which industry?Industry changes the answer more than anything else.
Median, average or range?An average is pulled up by a few high spenders.
What counts as IT spend?IT budget or all technology; capital or depreciation.
Is it a survey or a forecast?A market forecast is not a company budget.
Can you open the source?If you cannot read the method, treat the figure as low confidence.
Would you defend it to the CFO?If not, leave it out.

Give every comparison a confidence level

Not every comparison deserves the same weight. Rate each one before it reaches a slide:

ConfidenceWhenHow to use it
HighSame scope, your industry, recent data, a median or a range, method publishedEvidence: quote it with its source, date and statistic
MediumOne of these is weaker: an adjacent industry, older data, or a scope that needs adjustmentEvidence, with the caveat stated
LowSeveral are weak, or the method cannot be readContext only: show why it is set aside, never use it as evidence

This one discipline changes the conversation. Executives stop arguing about which survey is right and start asking what the reliable comparisons say.

Look beyond the ratio

The share of revenue is the least informative number in a good benchmark. Four views say more:

  • Your own trend over at least four years, on a constant scope. It is the most reliable comparison you have.
  • The mix: run against change, staff against external spend, capital against operating, security as a share of IT spend. Mix differences raise better questions than the level.
  • Business units: spend per employee or per unit of revenue by business unit, with corporate costs spread on an agreed key. Differences between your own units are easier to explain, and to act on, than differences with a survey.
  • Unit costs: cost per user, per device, per application, per service ticket. Compare them with last year and, where possible, with a few peers on agreed definitions.
Exchanging data with peers is often worth more than any survey: five or more companies of similar size and industry, comparing unit costs on agreed definitions. Because peers in the same industry are often competitors, set the rules before the first exchange: an independent third party collects and aggregates the data, no participant can be identified, historical costs only, no vendor prices or terms, no current or future plans or budgets, and legal advice first.

How to present it to the board

Keep it to a few slides, and lead with what you will do, not with the ratio.

  1. The level: total spend on the narrow and broad scopes, as a share of revenue, with the reliable comparisons and their confidence.
  2. The mix and the trend: where the money goes, and how it has moved.
  3. What a benchmark cannot tell you: whether the spend creates value, or what the right budget is for your strategy.
  4. What you will do: the questions the benchmark raised, the unit costs you will work on, and when you will report back.

Repeat the exercise once a year, before the budget round, on the same scope and rules. A benchmark refreshed every year becomes a trend; one done once becomes an argument.

Questions

What percentage of revenue should a company spend on IT?

There is no right percentage. It depends on the industry, the business model, the size of the company and its strategy. Use reliable comparisons for your industry to judge whether your level is plausible, then decide the budget from the strategy, the demand and the cost of each service.

Where can I find reliable IT spend benchmarks?

Research firms, industry associations and some public surveys publish figures. Prefer recent data for your industry, with a published method, sample and statistic, and record each figure with its source and date. Peer exchanges on unit costs are often more useful than any survey.

Should IT spend be compared on revenue or on operating cost?

Use revenue as the headline, because most published figures use it, but add spend per employee and as a share of operating expense. A business with high pass-through costs, such as a distributor, can look cheap on revenue and expensive against its own cost base, so state which denominator you use.

How often should we benchmark IT spend?

Once a year, before the budget round, on the same scope and rules, so that the trend is meaningful. A one-off benchmark is easy to dispute.

Toolkits on this topic

Related